weilingw/solubility-pipeline: Code used for Digital Discovery manuscript (with SI)

Machine learning pipeline for solubility prediction using XGBoost, SVM, and Random Forest models with molecular descriptors (MOE, RDKit, Mordred, Morgan fingerprints). Includes support for hybrid COSMO-RS features, SHAP-based interpretability, and substructure visualization. Developed for reproducible QSAR/QSPR studies in chemical informatics.